o .wi @sddlmZddlZddlmZddlmZdedededefd d Zdeded ed edef d dZ ddeded eedefddZ dS))OptionalN)Tensor)"_check_retrieval_functional_inputstargetpredsdiscount_cumsumreturnc Cstj| ddd\}}}tj|tjd}|d||j|jd||}|jddd}tj|tjd}||d|d<|||dd<|| S)aITranslated version of sklearns `_tie_average_dcg` function. Args: target: ground truth about each document relevance. preds: estimated probabilities of each document to be relevant. discount_cumsum: cumulative sum of the discount. Returns: The cumulative gain of the tied elements. T)return_inverse return_counts)dtyperdimN) torchunique zeros_likefloat32 scatter_add_tor cumsumdiffsum) rrr_invcountsrankedgroups discount_sumsrc/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/torchmetrics/functional/retrieval/ndcg.py_tie_average_dcgs  r top_k ignore_tiesc Cstdttj|jd|jdd}d||d<|r,|jdd}||}||}|S|jdd }t|||}|S) ayTranslated version of sklearns `_dcg_sample_scores` function. Args: target: ground truth about each document relevance. preds: estimated probabilities of each document to be relevant. top_k: consider only the top k elements ignore_ties: If True, ties are ignored. If False, ties are averaged. Returns: The cumulative gain g?)deviceg@gNT) descendingr ) rlog2arangeshaper$argsortrrr ) rrr!r"discountrankingrcumulative_gainrrrr_dcg_sample_scores-s$    r-cCst||dd\}}|dur|jdn|}t|tr|dks!tdt|||dd}t|||dd}|dk}d||<||||<|S) aCompute `Normalized Discounted Cumulative Gain`_ (for information retrieval). ``preds`` and ``target`` should be of the same shape and live on the same device. ``target`` must be either `bool` or `integers` and ``preds`` must be ``float``, otherwise an error is raised. Args: preds: estimated probabilities of each document to be relevant. target: ground truth about each document relevance. top_k: consider only the top k elements (default: ``None``, which considers them all) Return: A single-value tensor with the nDCG of the predictions ``preds`` w.r.t. the labels ``target``. Raises: ValueError: If ``top_k`` parameter is not `None` or an integer larger than 0 Example: >>> from torchmetrics.functional.retrieval import retrieval_normalized_dcg >>> preds = torch.tensor([.1, .2, .3, 4, 70]) >>> target = torch.tensor([10, 0, 0, 1, 5]) >>> retrieval_normalized_dcg(preds, target) tensor(0.6957) T)allow_non_binary_targetNr#rz,`top_k` has to be a positive integer or NoneF)r")rr( isinstanceint ValueErrorr-mean)rrr!gainnormalized_gainall_irrelevantrrrretrieval_normalized_dcgGsr6)N) typingrrrtorchmetrics.utilities.checksrr r0boolr-r6rrrrs   $